Skip to content

microsoft/power-platform-skills / plugins/power-pages/skills/edit-site

edit-site

Verified

Plans, coordinates and customizes PAC CLI-downloaded Standard or Enhanced data model non-SPA Power Pages sites created from Microsoft templates or manually in Design Studio. Use for requests involving pages, localized content, navigation, assets, snippets, site markers, Liquid/web templates, page layouts, or styling. Coordinated work produces an approved browser-viewable plan, executes changes in dependency order, verifies the local result, and optionally hands off to deployment. Not for React, Angular, Vue, or Astro SPA sites.

references/visual-asset-planning.md

View on GitHub
# Non-SPA Visual Asset Planning

Use this reference when a non-SPA site customization may benefit from photography, logos,
favicons, icons, illustrations, patterns, fonts, or other visual assets.

## Resolve delivery first

During site creation, use **direct HTTPS URLs for added images**, recorded as
`newSiteDesign.imageDelivery: "external-url"`. Do not download/stage images, create image Web Files
or attachment metadata, or execute an `author-web-file` operation for these additions. Reuse suitable existing
template images without reimporting them. Existing-site customization and explicit file imports
retain the Web File path below; this policy does not remove Web Files for CSS, fonts or documents.

Power Pages supports native [external image URLs](https://learn.microsoft.com/power-pages/getting-started/add-image).
For each external image:

- Resolve an actual direct image resource, not a gallery/HTML page or guessed photo ID. Require
  an absolute HTTPS URL without embedded credentials, raw whitespace or backslashes. Use a stable
  user-approved CDN/host URL or a curated Unsplash direct URL; never use auth/SAS secrets or
  expiring private links.
- Record `delivery: "external-url"`, `externalUrl`, source/license basis and
  `preparation: { "status": "remote" }`. Do not claim a file hash, cache path or imported metadata.
- Put that exact URL in the consumer's static `inputs` (including nested image `source` values).
  No import operation or `publicUrl` output binding is needed. For CSS backgrounds, also declare
  the URL in the styling owner's `externalResources`.
- Review hotlink permission, privacy and remote availability, plus known site CSP `img-src`
  restrictions. Power Pages permissions do not protect external URLs. Never relax CSP
  automatically; choose a permitted source or obtain separate approval for policy changes.
- Keep localized alt text, aspect ratio, reserved layout space, crop/focal point, and appropriately
  sized CDN variants. Source-page evidence and syntactic URL validation alone are not proof of
  availability. Keep unchecked external images explicitly **unverified**.

### Optional network checks

`render-edit-site-plan.js` makes no image-network requests by default. It
validates the plan and renders external images with an explicit unverified notice, without
creating an image-check report. Its result includes `imageChecks: null`, `verifiedImages: 0`,
and the number of unique unchecked URLs in `unverifiedImages`. Normal review and publication
can proceed in this state; neither a mandatory check-choice question nor a live site preview
is required.

Only when the user requests source verification, append `--verifyImages true` to the review
command. The existing verifier checks each unique `externalUrl` once before writing review HTML.
It uses up to three concurrent unauthenticated HTTPS GETs, a 20-second deadline per URL including
redirects, at most three redirects, and the existing 15 MiB asset byte limit.
Every destination must resolve to public Internet addresses; DNS answers are pinned for the
request. No cookies, auth headers, referrers, or image files are retained.

Require HTTP 200, a supported image Content-Type, non-empty bytes, and matching image content
signatures (PNG/APNG, JPEG, WebP, GIF, AVIF, SVG, or ICO). This is an HTTP/content check, not a full
browser decode or visual-fidelity guarantee. For oversized assets, choose a smaller CDN variant.
A requested check's 404, HTML/gallery response, timeout, or other failure blocks review generation with the affected
asset names and URLs. Select a working direct resource and update both its manifest entry and
all consumer inputs before rendering again; never repeatedly retry a known 404 or substitute
an unapproved image after approval. Do not disable checking to hide a known failure or silently
fall back to unchecked HTML.

Successful checks are stored in `plan.html.image-checks.json`, bound to the complete plan hash
and exact URLs, and displayed in the review. If the approved review included checks, publish with
`--imageChecks` pointing to that report; otherwise omit the argument. Supplied reports must
validate and are preserved in `current-execution.json`; `--action resolve` supplies them
to the executing workflow without more requests. Missing, stale or failed supplied reports never silently become
unchecked publication. Reuse valid evidence within that approved run. A revised plan requires
fresh review and approval, with fresh checks only when requested. The timestamps describe
observed availability, not a promise that remote content cannot change. Receipts without reports,
including legacy receipts, remain resumable without claiming checks that were not performed.

The generated plan lazily displays approved hosted images in the browser with localized preview
alt text, the planned aspect ratio, and a no-referrer policy. Retain clickable image/source links,
known attribution, license, and hosting caveats alongside the preview. If an image cannot load, show an
explicit unavailable state with an image-source link, not a broken-image box or endless loading
message. Pure HTML serialization and publication do not fetch images; an optional source check
reads them in memory without staging files or importing Web Files.
Browser preview requests go to the approved provider; successful previews do not prove site CSP,
final crop/focal-point treatment, or Studio/runtime rendering. A local-only or newly authored image
without an approved hosted URL retains an explicit unavailable preview and its file metadata; do
not upload it, guess a runtime host, or invent a URL to manufacture a preview.

## Design role before source

Do not add an asset merely to fill a hero, card, or empty column. Inspect the visitor journey and
classify each proposed visual by the job it performs:

| Role | Typical assets | Planning rule |
|---|---|---|
| Brand | Logo, wordmark, favicon, brand pattern | Prefer existing or user-approved brand files |
| Informative | Product image, map, diagram | Require a source that communicates the stated fact |
| Editorial | People, place, service photography | Use only when it improves credibility or context |
| Structural | Section divider, background motif | Prefer original SVG or CSS atmosphere |
| Functional | Icons, state markers, directional cues | Use a coherent original SVG set and never rely on color alone |
| Decorative | Abstract flourishes and textures | Keep restrained and use empty alternative text |

FAQ, pricing, policy, and operational pages often benefit more from typography, spacing, and
simple original SVG details than unrelated stock photography. Record `assets: []` when no visual
asset improves an existing-site or narrow-edit outcome.

For an explicit new-site design handoff, plan meaningful content imagery as part of the initial
composition: a relevant hero image or original illustration and supporting images where they
explain the content. Reuse suitable template images rather than replacing them gratuitously.
Suitability means fit with the approved experience, subject and art direction, not merely presence
in the selected creation template. New-site imagery and branding are not constrained by that
template; replace unsuitable visuals through the approved asset/caller operations without
discarding required content or deleting unrelated Web Files. Honor explicit asset-preservation
preferences and keep existing-site narrow edits within their requested scope.
Record at least one non-decorative content photograph/illustration in the asset manifest with
its real placement; a logo, favicon, icon, or decorative pattern alone is insufficient. An explicit
user request to omit images is recorded as `newSiteDesign.imagery: "user-declined"` with
`imageryReason`. Missing source material is a dependency to resolve, not an implicit opt-out.
Use the shared `${PLUGIN_ROOT}/references/site-design-quality.md` for visual composition; this
reference owns the delivery-specific URL, staging/import, binding, and verification rules.

## Source priority

Among sources that fit the approved design, resolve in this order:

1. existing approved site/template images or hosted image URLs;
2. user-provided approved hosted brand assets and photography;
3. curated Unsplash photography for URL-based additions.

Outside URL-based creation, user-provided local files and original safe SVG illustrations may
also use the explicitly selected Web File delivery path. Do not silently switch a new-site image
to file delivery because a hosted source is missing.

Do not generate raster images. Do not use another stock provider. Do not use placeholder-image
services or leave unresolved image slots.

### Existing site assets

Inspect `web-files/`, image callers, logo/favicon callers, snippets, templates, and comparable
pages. Reuse an existing asset only when its identity, public URL, permissions boundary, visual
role, and licensing/ownership are appropriate for the new placement.

### User-provided assets

This is the Web File delivery path only, not the new-site URL image path. Treat every supplied
file as untrusted. Stage it with:

```bash
node "${PLUGIN_ROOT}/scripts/prepare-non-spa-asset.js" \
  --projectRoot "<PROJECT_ROOT>" \
  --sourcePath "<SOURCE_FILE>"
```

The helper validates supported signatures, dimensions, MIME agreement, size, and SVG safety,
writes approved bytes under `.powerpages-customization/assets/`, adds that cache to `.gitignore`,
and returns the project-relative path and SHA-256. The cache is outside the non-SPA site root
and must never be passed to `pac pages upload`.

### Original SVG assets

Use local SVG authoring/staging only on the explicit Web File path. URL-based creation needs an
already hosted approved illustration; do not stage or upload an illustration as a hidden fallback.

Original SVG is suitable for simple geometry, icons, patterns, and non-photographic illustration.
Author it in the fresh external review directory, then stage it through the same helper before the
plan is approved. Use a real `viewBox`; do not encode visitor-facing text into the SVG when HTML
text would be more accessible and localizable.

The helper rejects declarations/entities, external stylesheets, script or style elements, style
and event-handler attributes, embedded HTML or media, and external resource references. Use SVG
presentation attributes for simple geometry. Do not weaken those checks to accommodate a design.

### Unsplash photography

Follow the same **WebSearch-to-CDN** sourcing approach as SPA creation:

1. Use `WebSearch` for the specific subject the section needs. Choose photography whose
   composition, lighting, color temperature, orientation, and available text space fit the page.
2. Use a direct `https://images.unsplash.com/photo-...` URL found through the search, with explicit
   sizing/crop parameters such as `w=1600&h=900&fit=crop&fm=jpg` (`w=800` for cards, `w=1600` for
   heroes). Preserve the discovered CDN path. Do not derive it from a photo-page slug or a
   search-result identifier, guess IDs, or copy documentation placeholders.
3. Keep one art direction across the chosen photos, with consistent crops and localized alt text.
   Record the exact sized/cropped URL as `externalUrl`, `source.type: "unsplash"`, the
   `source.license` basis, and `preparation: { "status": "remote" }`.

For external delivery, **do not require a photo-page fetch, scraping, API credentials, or a
photographer lookup**. Retain `source.sourcePage` and `source.photographer` when reliably known
from search results or supplied evidence; otherwise omit them rather than inventing attribution
or a photo-page/CDN pairing. Supplied metadata must still be valid.

An anti-bot challenge or unavailable `unsplash.com` photo/search page is not evidence that
`images.unsplash.com` is unavailable. Do not bypass the challenge or repeatedly retry it;
continue with a direct CDN URL found through WebSearch. If none is found, search for another
suitable Unsplash photo or request an approved hosted URL. Continue content/layout planning
while resolving imagery, but do not request approval with unresolved images.

Do not download the image to a file. Network verification is optional as described above;
when requested, check the **final direct image URL**, not the photo page. A failed CDN image
check still blocks approval HTML for that checked review; successful checks remain bound to
the exact plan and are published through `--imageChecks`. Without checks, preserve the explicit
unverified status.

For an explicitly selected Web File delivery outside that creation path, retain the required
photo page, photographer, license and `source.downloadUrl`, then stage the selected image:

```bash
node "${PLUGIN_ROOT}/scripts/prepare-non-spa-asset.js" \
  --projectRoot "<PROJECT_ROOT>" \
  --downloadUrl "<HTTPS_IMAGES_UNSPLASH_URL>" \
  --sourcePage "<HTTPS_UNSPLASH_PHOTO_PAGE>" \
  --fileName "<SAFE_FILENAME_WITH_EXTENSION>"
```

[Unsplash's published license](https://unsplash.com/license) permits downloading, copying, modifying, distributing, and
commercial use without required attribution. The license does not grant separate rights for
recognizable people, trademarks, artwork, or protected property visible in a photograph. Avoid
those subjects when the planned use could imply endorsement or require additional rights; retain
a plan warning when the context needs human legal/brand review.

Import staged files as local Power Pages Web Files only on that file-delivery path. URL-based
creation uses the approved Unsplash direct URL in the finished image component instead.

## Composition decisions

For each approved asset, resolve:

- role and visitor-facing purpose;
- page, section, and global-versus-page scope;
- visual rationale;
- aspect ratio, crop, and focal-point intent;
- informative versus decorative behavior;
- localized alternative text for every affected locale;
- optional safe link destination;
- existing, remote, staged, or global-caller dependencies;
- whether `style-site` is required for presentation after placement.

Use the image component's responsive defaults unless the inspected site establishes a stronger
pattern. Do not invent unsupported image components, inline arbitrary data URIs, or copy Design
Studio placeholder SVG payloads.

## Logos and favicons

During every header update, match the logo or wordmark to the approved website name and brand
requirements. Do not preserve an unrelated starter logo merely because it came with the template.
Do not silently invent an official company logo or use stock photography as a substitute.
Resolve the direction using known inputs:

1. use a supplied approved logo when the requirements call for it;
2. retain the verified current logo only when it meets those requirements or the user explicitly
   requests preservation, and record why;
3. if no suitable logo is available, propose a clearly labeled typography-only wordmark or simple
   original geometric mark for approval in the existing customization plan;
4. when brand authorization or required source material is unresolved, ask only for that missing
   decision in the normal clarification round. Report unresolved logo work as incomplete rather
   than silently choosing no change. Honor an explicit no-logo or no-change request.

A typography-only wordmark uses native text/HTML. Image marks still follow the resolved delivery
contract: new-site additions need an approved hosted URL, not an automatic upload or a new Web File.
Keep logo and meaningful content imagery separate; a logo alone does not satisfy the new-site
imagery requirement.

A logo or favicon change is site-wide even when only one Web File is imported. Resolve the exact
existing snippets, templates, site settings, or other callers, including separate desktop/mobile
logo variants and requested locales. Update only those approved callers, with the correct brand
name/alternative text, home-link destination, aspect ratio and responsive sizing intent. Do not
silently expand a header-logo change to favicons, footer branding or unrequested locales.
Do not replace the complete website header merely to change a logo, and preserve navigation,
search, language selection, sign-in behavior, caching assumptions, and anonymous/authenticated
branches.

## Execution and verification

External images require no import: consumers use their approved static HTTPS URL directly.
Verify that actual image references, after HTML attribute decoding, match that URL and that no
download/cache/Web File was created. Do not treat a remote asset as a byte-hashed local artifact.

For Web File delivery, create/import operations use the `author-web-file` reference workflow
from `${PLUGIN_ROOT}/skills/edit-site/references/web-file-authoring.md`.
Their consumers depend on
the import and bind the verified `publicUrl`; never guess a future Web File URL.

For Web File assets, after execution verify:

- staged SHA-256 matches the imported bytes;
- binary and adjacent `.webfile.yml` both exist;
- extension, MIME type, filename, metadata, IDs, parent, and public URL agree;
- every planned caller uses the final root-relative URL;
- informative images have localized alternative text and decorative images use `alt=""`;
- dimensions preserve aspect ratio and fit the approved Design Studio column;
- no unapproved SVG active content or base64 placeholder remains;
- global brand callers and unrelated assets are preserved;
- target-environment attachment size and blocked-extension policy remain satisfied.

Local inspection cannot prove the final crop, visual balance, CSP behavior, or live rendering.
Keep desktop/mobile crop, readability over imagery, logo scaling, and runtime performance as
pending live checks unless they are separately authorized and performed.
writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deploymentStars: 322kUpdated Claude CodeCodexVerifiedInstall
xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like "the xlsx in my downloads") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy.Stars: 119kUpdated Claude CodeCodexVerifiedInstall
childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release.Stars: 118kUpdated Claude CodeCodexVerifiedInstall